Exposure
A prerequisite for the correct interpretation of effects of chemicals on biota is the confirmation of intended exposure regimes in experimental studies. Table 1 lists the measured concentrations of chlorpyrifos at the start of the experiments. The intended concentrations were well achieved with an average coefficient of determination of 0.97 ± 0.04, average slope = 0.86 ± 0.22, and average intercept = –0.55 ± 2.77 for linear regression across all experiments of intended and measured concentrations at t = 0. As expected, during the course of the experiment, the measured concentrations of chlorpyrifos in most of the test media decreased (Table 2). The experiments with the different species, however, showed differences in dissipation of chlorpyrifos ranging from 55.2% to 111.7% remaining after 48 h exposure and 21.9% to 120.7% remaining after 96 h of exposure (Table 2). Although no consistent pattern could be found when comparing species or treatments, the interplay of factors, such as animal size, bioconcentration, evaporation, and degradation, could explain these differences. For further calculation of L(E)C50 values and SSDs, the initial concentrations in the static systems were used because of the short test duration and the relatively long organism recovery time shown for organophosphates (Ashauer et al. 2007).
Table 2 Dissipation of chlorpyrifos in test media during the course of the experiment relative to initially measured concentrations
In general, concentrations of chlorpyrifos in control replicates were lower than the respective LODs, but occasionally chlorpyrifos was measured in single control samples (C. obscuripes, D. magna, N. denticulata, P. stratiotata, Procambarus spec. juveniles, and S. lutaria), which explains the high variability in the intercept reported previously. These exceptions are related to cross-contamination of controls with chlorpyrifos due to its high volatility and associated contamination routes, especially when experiments with high-exposure concentrations were performed. In all the controls showing cross-contaminations, immobilization or mortality was < 10% and therefore tolerated in the presented study. In the experiment with M. angustata, all control replicates were contaminated on average with 0.143 μg/l chlorpyrifos, and immobility was induced in one to two control animals per replicate (20% to 40%), leading indirectly to cannibalism. As a result, control mortality and increased mortality in the lower concentrations was observed during the experiment. Cannibalism was also observed at the lower concentrations, but not in the intermediate and high concentrations, where no mortality but full immobilisation was induced in all test animals and thus prevented cannibalism. Due to a current lack of data on the effects of chlorpyrifos for this species, it was decided rather to correct the total number of immobile or dead animals for further analysis instead of excluding the species. The data correction was performed by setting both concentrations of the contaminated control replicates and the induced immobility/mortality at 24 h in the first two treatments to zero.
Effects on Mortality and Immobility
Effects of chlorpyrifos induced in a range of freshwater arthropods by short-term exposure in simple laboratory test systems are presented as concentration response relations for mortality and immobility (Tables 3 and 4). Respective concentration–response parameters are reported (according to Equation 1) next to L(E)C50 values and their confidence limits. At least the highest test concentrations of chlorpyrifos induced 100% immobility within the 96-h exposure in all species except in C. dipterum, where 93% of all test animals were immobile at the end of the experiment. In P. stratiotata, all initial test animals in the highest concentration were immobile at 72 h of exposure, but subsequent recovery occurred, resulting in 70% immobilization at the end of the experiment. In contrast, at least the highest concentrations induced 100% mortality only in 6 of the 14 experiments (Anax imperator, G. pulex, P. minutissima, Procambarus spec. adults and juveniles, and R. linearis), and for some species no mortality (S. lutaria, M. angustata) or low mortality (A. aquaticus) was observed within the time of exposure, even at the highest concentrations. For the remaining 5 species, we found between 70% and 87% mortality (C. obscuripes, C. dipterum, D. magna, N. denticulata, Notonecta maculata, and P. stratiotata) at their respective highest concentrations. Figure 1 illustrates that for some species the differences between lethal and sublethal effect concentrations are substantial independent of time, whereas others show a relatively good match.
Table 3 Lethal effects of chlorpyrifos on freshwater arthropods in 96 ha
Table 4 Sublethal effects of chlorpyrifos (immobility) on freshwater arthropods in 96 ha
The observed effects on mortality and immobilisation for the tested species is in the range reported in literature, which is up to 3 to 4 orders of magnitude for both lethal and sublethal effects for up to 96 h (Van Wijngaarden et al. 1993; Maltby et al. 2005; Rubach et al. 2010). In addition, L(E)C50 values for particular species agree well with previous findings, with the exception of C. obscuripes, for which an LC50 22 times higher (6.6 μg/l, 96-h exposure) was previously determined (Van Wijngaarden et al. 1993). These investigators also observed differences in the concentrations inducing mortality and immobility for some arthropod taxa (A. aquaticus, Proasellus coxalis, G. pulex, C. dipterum, C. horaria, and C. obscuripes) and, again, no mortality was induced within 96 h of exposure for A. aquaticus and also Caenis horaria in their study. Although the experiments of Van Wijngaarden et al. (1993) were either performed in flow-through or semistatic test systems and therefore under constant exposure, the congruence of results with our study shows that for chlorpyrifos exposures up to 96 h the initial (peak) concentrations are equally representative for effects induced within this period of time.
Dynamics of Effects and Their Variability
As expected, the effects of chlorpyrifos on both mortality and immobility increase in time in all tested species, which can be seen from the overall decrease in L(E)C50 values (Fig. 2 and Tables 3, 4). Chlorpyifos mainly acts on the nervous system by inhibiting the enzyme acetylcholinesterase, leading to a synaptic block and therefore inhibiting electric signal transmission. This first leads to sublethal intoxication symptoms, and subsequent death is likely caused by final respiratory failure (Eaton et al. 2008). Hence, as expected, observations of immobility consistently resulted in lower EC50 values in time compared with their respective LC50 values, although the difference between these end points decreased during the course of the experiment, especially for D. magna, for which the LC50/EC50 ratios decreased from 128.6 to 4.75 in 72 h (Fig. 2). In general, it is logical that the effect concentrations for immobility and mortality will converge to the same value with time; however, it is evident that this does not occur with the same speed for all the tested species (Fig. 2). For some species, the differences between LC50 and EC50 even stayed relatively constant within the 96 h of test duration. However, for the species A. imperator, C. dipterum, G. pulex, Procambarus spec., and R. linearis, a good match between effective and lethal concentrations was observed right from the start of the experiments (estimated LC50/EC50 ratios approximately 1; see also Figs. 1 and 2). In contrast, for a third group of species (N. denticulata, N. maculata, and P. stratiotata), the difference between EC50 and LC50 values for a particular species does not necessarily change in time (LC50/EC50 ratios were constantly approximately 2, 2, and 9, respectively). In addition, the extent to which LC50 and EC50 values differ for certain time points seems rather species-specific, especially for S. lutaria and M. angustata, in which no significant incipient mortality was induced by the applied concentrations but in which immobility was induced at quite low concentrations (Fig. 2). This is interesting in the sense that these species-specific differences in incipient mortality or immobility can be either due to differences in toxicokinetics and/or toxicodynamics. For instance, on one hand, S. lutaria, M. angustata, and A. aquaticus could have the ability to either decrease or regulate uptake and/or elimination of chlorpyrifos, to biotransform chlorpyrifos slower to the chlorpyrifos-oxon, or to detoxify the latter quickly and therefore delay incipient mortality significantly, all of which would relate to differences in toxicokinetics. In contrast, differences in the species responses might be caused by other processes pertaining to toxicodynamcis, e.g., differences in the interaction of chlorpyrifos and acetylcholinesterase (target enzyme) or in the ability to compensate or repair damage. For details on toxicokinetics and toxicodynamics see Ashauer et al. (2006). Rubach et al. (in press) measured uptake and elimination kinetics of 14C-labeled chlorpyrifos in the same species and indicated that ≤38% of the variation in sensitivity (EC50, immobilisation in 48 h, same data) may be explainable by uptake and ≤28% by elimination kinetics. Interestingly, S. lutaria, A. aquaticus, and M. angustata, which responded with a remarkable concentration difference between incipient immobility and mortality in this study, show high bioconcentration factors (9625, 3242, and 5331 μg/kgww, respectively). Because their uptake rates are moderate to high, and because immobility is effectively induced at much lower concentrations, differences in uptake itself can be excluded. More likely are differences in biotransformation rates (either bioactivation or detoxification) or a highly efficient compensatory gene-regulation ability. The most insensitive of the investigated species, N. denticulata, shows high uptake and high elimination rates and therefore a moderate bioconcentration, which partly explains its insensitivity.
Clearly, the extent of variation in observed sensitivity to chlorpyrifos across species highly depends on the end point under consideration, which is already evident from the concentration–response relations but also from the SSDs shown in Fig. 3. The SSDs also indicate by the “left shift” of both mortality and immobilisation that effects increase in time. The slopes of the SSDs for immobility do not seem to be significantly different; however, the variability increases slightly in time, as indicated by an increase in σ’ (Table 5). Nevertheless, the CIs of the SSD (Table 5) show a strong overlap; therefore, this trend cannot be confirmed reliably, and species variation in immobilisation might still be relatively constant in time. The variation in mortality is generally much higher and also relatively constant in time (σ’, Table 5) until 72 h of exposure, after which a sharp increase in species variability was observed. This is evident from the high σ’ (Table 5) and low slope of the SSD for 96 h, but only if all available species are included in the SSDs (Fig. 3). This impact on the slope is an artefact caused by the species selection and the close-to-zero mortality in A. aquaticus, M. angustata, and S. lutaria, for which no LC50 could be calculated and which were thus not included in the SSDs for exposure times ≤96 h. If these insensitive species (with high LC50 values) had been included for all time points, the slopes of these mortality SSDs would have been similar in time or even lower than the one for 96 h of exposure, and σ’ would have indicated an even bigger variation. Therefore, the variability in immobility is generally lower than for mortality; however, both are relatively stable in time if based on the same selection of species. The ratio of HC50 (mortality)/HC50 (immobility) derived from Table 5 decreases slowly in time (2.9, 2.4, 2.1, 2.0) when based on the same species selection; similarly, if all 14 species are included in the 96-h mortality SSD this ratio is 4.4. This shows that strong differences in time-dependent toxicity exist between species and that the SSD can only account for this if species selection is restricted to similarly reacting groups.
Table 5 SSD characteristicsa
Choice of End Point for ERA
Until now, from the perspective of individual sensitivity response, the presented data support the assumption that immobility is the better end point when investigating a neurotoxic substance, such as chlorpyrifos, because paralysis is the first visible symptom, and a species-specific time lag until incipient mortality was found for several species. In addition, also from a population sustainability viewpoint, immobility is almost just as relevant as mortality. Where on one hand, mortality is unilateral in the sense that a dead specimen cannot become alive again, any immobile or otherwise sublethally affected specimen may become mobile again and thus be able to contribute to a population’s sustainability. In contrast, an immobile specimen is likely going to be outcompeted, starved, drifted, or predated quickly under field conditions and also more prone to multiple stress regimes.
However, when the risk assessment is based on SSDs rather than on safety factors, including only data based on immobility, illogically does not always yield the most conservative threshold but may deliver a more confident estimate of the HC5 due to less variation in the species selection. Steep SSDs with higher confidence, as calculated here for immobility, will deliver less conservative HC5 values than shallower SSDs, such as shown here for mortality after 96 h of exposure. This is especially the case if the lower end of the curve shows a bad fit with the data (as the 96-h mortality SSD for all species in Fig. 3). Such SSDs can, however, be excluded using goodness-of-fit measures, especially the Anderson–Darling test, which is sensitive for the quality of fit in the lower concentration range (Van Vlaardingen et al. 2004). All presented SSDs passed the three performed goodness-of-fit tests with p < 0.001, except for the 96-h mortality SSD with all species included, which did not pass the Anderson–Darling and the Cramer van Mises tests (both p < 0.1). The lower confidence limit of the HC5 (LLHC5) derived from an SSD may serve as a protective threshold in higher-tier risk assessment (Maltby et al. 2009; Brock et al. 2010). This value can differ substantially between different observation times and different end points (see Table 5) depending on the species selection. Although in general a rather conservative threshold, the protectiveness of the LLHC5 highly depends on the incipient time of the effect, the end point included (which correlates to the incipient of effects), the extent to which the species included in the SSD vary in their sensitivity, and other quality criteria as reviewed in Brock et al. (2010). Our results show that the most confident estimates can be derived with an SSD when immobility after sufficient time of exposure is chosen as an end point for the SSD. Herewith, if these “quality criteria” are addressed, the convolution caused by inclusion of insensitive species does not touch the usefulness of the SDD as a tool for ERA, especially if the most sensitive group for one chemical is well represented in the SSD (Van Den Brink et al. 2006); however, it also clearly shows that an SSD does not necessarily represent the true existing variation in sensitivity.
Other end points, in addition to mortality and immobility, describing population sustainability, such as reproduction, can be derived from chronic testing but cannot be deduced from short-term tests. To derive a good and conservative proxy for population sustainability, another sublethal end point such as postexposure feeding inhibition should be considered for short-term testing (McWilliam and Baird 2002; Satapornvanit et al. 2009). If a specimen is not able to feed within a given time period, e.g., 24 h after the end of a short-term exposure, it is rather unlikely that it is able to contribute to the population’s sustainability. This, however, is yet far from being taken into account in current risk-assessment practices. Another problematic issue for the selection and definition of an appropriate end point for ERA when comparing effects on species are the criteria that must be set for this particular end point. For instance, in this study, transparent species could be observed for heartbeat and thus had a good criterion by which to distinguish death from immobility. Nontransparent and highly-sclerotised species, however, do not provide such clear-cut criteria to determine clinical death. In the present study, the end point criteria for each species were rather well defined, thus minimizing such described difficulties.
To improve ERA, the identification of the best end point for assessing the risk of a certain group of chemicals must be based on its exposure scenario, its functional relevance for the mode or mechanism of action, its toxicity in time or on other previous knowledge, and the ecologic consequences of a given end point.